Final Project: Oxford-IIIT Pet Dataset
Bilawal Shaikh and Faizan Dogar
In this project, we will be training on the Oxford-iiit-pets dataset which has 37 categoies: 25 categories of dogs and 12 categories of cats. We pass our images through the ResNet50 model (without the final output classification layer).
We start by loading libraries relevant to our project.
library(readr)
library(ggplot2)
library(dplyr)
library(methods)
library(stringi)
library(keras)We now load the ResNet50 model and grab the second last layer from the model.
resnet50 <- application_resnet50(weights = 'imagenet', include_top = TRUE)
model_embed <- keras_model(inputs = resnet50$input,
outputs = get_layer(resnet50, 'avg_pool')$output)We load all of our data from our csv and rds files which were constructed in another class:
pets <- read_csv("my-image-data.csv")## Parsed with column specification:
## cols(
## obs_id = col_character(),
## train_id = col_character(),
## class = col_double(),
## class_name = col_character(),
## path_to_image = col_character()
## )x69 <- read_rds("final_project.rds")We now produce our embeddings to get our X and y matrices:
X <- t(apply(x69, 1, cbind))
y <- pets$class
class_names <- levels(factor(pets$class_name))We create our training dataset from our matrices, where the training/validation split is 60/40 respectively:
X_train <- X[pets$train_id == "train",]
y_train <- to_categorical(pets$class[pets$train_id == "train"])To see a sample of some of the images in the dataset:
set.seed(1)
par(mfrow = c(2, 2))
samples = sample(0:37, 5)
for (j in samples) {
id <- sample(which(y == j), 4)
for (i in id) {
par(mar = rep(0, 4L))
plot(0,0,xlim=c(0,1),ylim=c(0,1),axes= FALSE,type = "n")
Z <- image_to_array(image_load(pets$path_to_image[i], target_size = c(224,224)))
rasterImage(Z /255,0,0,1,1)
text(0.5, 0.1, label = class_names[y[i] + 1L], col = "red", cex=2)
}
}
We create a keras model with multiple layers added:
model <- keras_model_sequential()
model %>%
layer_dense(units = 256, input_shape = ncol(X_train)) %>%
layer_activation(activation = "relu") %>%
layer_dropout(rate = 0.5) %>%
layer_dense(units = 256) %>%
layer_activation(activation = "relu") %>%
layer_dropout(rate = 0.5) %>%
layer_dense(units = ncol(y_train)) %>%
layer_activation(activation = "softmax")We compile our model:
model %>% compile(loss = 'categorical_crossentropy',
optimizer = optimizer_rmsprop(lr = 0.001 / 2),
metrics = c('accuracy'))We fit our model to our dataset:
history <- model %>% fit(X_train, y_train, epochs = 20)
plot(history)
We find our predictions from this model:
y_pred <- predict_classes(model, X)
tapply(y == y_pred, pets$train_id, mean)## train valid
## 0.9944367 0.9036857We see that the model overfits the data (99.4% training accuracy) but the validation accuracy is fairly high at 90.2%.
We also print the confusion matrix:
table(value = class_names[y + 1L], prediction = class_names[y_pred + 1L], pets$train_id)## , , = train
##
## prediction
## value Abyssinian American bulldog
## Abyssinian 120 0
## American bulldog 0 118
## American Pit Bull Terrier 0 0
## Basset Hound 0 0
## Beagle 0 0
## Bengal 0 0
## Birman 0 0
## Bombay 0 0
## Boxer 0 1
## British Shorthair 0 0
## Chihuahua 0 0
## Egyptian Mau 0 0
## English Setter 0 0
## German shorthaired 0 0
## Great Pyreness 0 0
## Havanese 0 0
## Japanese Chin 0 0
## Keeshond 0 0
## Leonberger 0 0
## Maine Coon 0 0
## Miniature Pinscher 0 0
## Newfoundland 0 0
## Persian 0 0
## Pomeranian 0 0
## Pug 0 0
## Ragdoll 0 0
## Russian Blue 0 0
## Saint Bernard 0 0
## Samoyed 0 0
## Scottish Terrier 0 0
## Shiba Inu 0 0
## Siamese 0 0
## Sphynx 0 0
## Staffordshire Bull Terrier 0 0
## Wheaten Terrier 0 0
## Yorkshire Terrier 0 0
## prediction
## value American Pit Bull Terrier Basset Hound Beagle
## Abyssinian 0 0 0
## American bulldog 1 0 0
## American Pit Bull Terrier 118 0 0
## Basset Hound 0 120 0
## Beagle 0 2 118
## Bengal 0 0 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 0 0 0
## Egyptian Mau 0 0 0
## English Setter 0 0 0
## German shorthaired 0 0 1
## Great Pyreness 0 0 0
## Havanese 0 0 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 0 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 0 0
## Shiba Inu 0 0 0
## Siamese 0 0 0
## Sphynx 0 0 0
## Staffordshire Bull Terrier 1 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value Bengal Birman Bombay Boxer British Shorthair
## Abyssinian 0 0 0 0 0
## American bulldog 0 0 0 0 0
## American Pit Bull Terrier 0 0 0 0 0
## Basset Hound 0 0 0 0 0
## Beagle 0 0 0 0 0
## Bengal 120 0 0 0 0
## Birman 0 120 0 0 0
## Bombay 0 0 120 0 0
## Boxer 0 0 0 119 0
## British Shorthair 0 0 0 0 120
## Chihuahua 0 0 0 0 0
## Egyptian Mau 1 0 0 0 0
## English Setter 0 0 0 0 0
## German shorthaired 0 0 0 0 0
## Great Pyreness 0 0 0 0 0
## Havanese 0 0 0 0 0
## Japanese Chin 0 0 0 0 0
## Keeshond 0 0 0 0 0
## Leonberger 0 0 0 0 0
## Maine Coon 0 0 0 0 0
## Miniature Pinscher 0 0 0 0 0
## Newfoundland 0 0 0 0 0
## Persian 0 0 0 0 0
## Pomeranian 0 0 0 0 0
## Pug 0 0 0 0 0
## Ragdoll 0 8 0 0 0
## Russian Blue 0 0 0 0 1
## Saint Bernard 0 0 0 0 0
## Samoyed 0 0 0 0 0
## Scottish Terrier 0 0 0 0 0
## Shiba Inu 0 0 0 0 0
## Siamese 0 0 0 0 0
## Sphynx 0 0 0 0 0
## Staffordshire Bull Terrier 0 0 0 0 0
## Wheaten Terrier 0 0 0 0 0
## Yorkshire Terrier 0 0 0 0 0
## prediction
## value Chihuahua Egyptian Mau English Setter
## Abyssinian 0 0 0
## American bulldog 0 0 0
## American Pit Bull Terrier 0 0 0
## Basset Hound 0 0 0
## Beagle 0 0 0
## Bengal 0 0 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 119 0 0
## Egyptian Mau 0 119 0
## English Setter 0 0 120
## German shorthaired 0 0 0
## Great Pyreness 0 0 0
## Havanese 0 0 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 0 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 0 0
## Shiba Inu 0 0 0
## Siamese 0 0 0
## Sphynx 0 0 0
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value German shorthaired Great Pyreness Havanese
## Abyssinian 0 0 0
## American bulldog 0 0 0
## American Pit Bull Terrier 0 0 0
## Basset Hound 0 0 0
## Beagle 0 0 0
## Bengal 0 0 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 0 0 0
## Egyptian Mau 0 0 0
## English Setter 0 0 0
## German shorthaired 119 0 0
## Great Pyreness 0 120 0
## Havanese 0 0 119
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 0 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 0 0
## Shiba Inu 0 0 0
## Siamese 0 0 0
## Sphynx 0 0 0
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value Japanese Chin Keeshond Leonberger Maine Coon
## Abyssinian 0 0 0 0
## American bulldog 0 0 0 0
## American Pit Bull Terrier 0 0 0 0
## Basset Hound 0 0 0 0
## Beagle 0 0 0 0
## Bengal 0 0 0 0
## Birman 0 0 0 0
## Bombay 0 0 0 0
## Boxer 0 0 0 0
## British Shorthair 0 0 0 0
## Chihuahua 0 0 0 0
## Egyptian Mau 0 0 0 0
## English Setter 0 0 0 0
## German shorthaired 0 0 0 0
## Great Pyreness 0 0 0 0
## Havanese 0 0 0 0
## Japanese Chin 119 0 0 0
## Keeshond 0 120 0 0
## Leonberger 0 0 120 0
## Maine Coon 0 0 0 120
## Miniature Pinscher 0 0 0 0
## Newfoundland 0 0 0 0
## Persian 0 0 0 0
## Pomeranian 0 0 0 0
## Pug 0 0 0 0
## Ragdoll 0 0 0 0
## Russian Blue 0 0 0 0
## Saint Bernard 1 0 0 0
## Samoyed 0 0 0 0
## Scottish Terrier 0 0 0 0
## Shiba Inu 0 0 0 0
## Siamese 0 0 0 0
## Sphynx 0 0 0 0
## Staffordshire Bull Terrier 0 0 0 0
## Wheaten Terrier 0 0 0 0
## Yorkshire Terrier 0 0 0 0
## prediction
## value Miniature Pinscher Newfoundland Persian
## Abyssinian 0 0 0
## American bulldog 0 0 0
## American Pit Bull Terrier 0 0 0
## Basset Hound 0 0 0
## Beagle 0 0 0
## Bengal 0 0 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 1 0 0
## Egyptian Mau 0 0 0
## English Setter 0 0 0
## German shorthaired 0 0 0
## Great Pyreness 0 0 0
## Havanese 0 1 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 120 0 0
## Newfoundland 0 120 0
## Persian 0 0 119
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 0 0
## Shiba Inu 0 0 0
## Siamese 0 0 0
## Sphynx 0 0 0
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value Pomeranian Pug Ragdoll Russian Blue
## Abyssinian 0 0 0 0
## American bulldog 0 0 0 0
## American Pit Bull Terrier 0 0 0 0
## Basset Hound 0 0 0 0
## Beagle 0 0 0 0
## Bengal 0 0 0 0
## Birman 0 0 0 0
## Bombay 0 0 0 0
## Boxer 0 0 0 0
## British Shorthair 0 0 0 0
## Chihuahua 0 0 0 0
## Egyptian Mau 0 0 0 0
## English Setter 0 0 0 0
## German shorthaired 0 0 0 0
## Great Pyreness 0 0 0 0
## Havanese 0 0 0 0
## Japanese Chin 1 0 0 0
## Keeshond 0 0 0 0
## Leonberger 0 0 0 0
## Maine Coon 0 0 0 0
## Miniature Pinscher 0 0 0 0
## Newfoundland 0 0 0 0
## Persian 0 0 1 0
## Pomeranian 120 0 0 0
## Pug 0 120 0 0
## Ragdoll 0 0 112 0
## Russian Blue 0 0 0 119
## Saint Bernard 0 0 0 0
## Samoyed 0 0 0 0
## Scottish Terrier 0 0 0 0
## Shiba Inu 0 0 0 0
## Siamese 0 0 0 0
## Sphynx 0 0 0 0
## Staffordshire Bull Terrier 0 0 0 0
## Wheaten Terrier 0 0 0 0
## Yorkshire Terrier 0 0 0 0
## prediction
## value Saint Bernard Samoyed Scottish Terrier
## Abyssinian 0 0 0
## American bulldog 0 0 0
## American Pit Bull Terrier 0 0 0
## Basset Hound 0 0 0
## Beagle 0 0 0
## Bengal 0 0 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 0 0 0
## Egyptian Mau 0 0 0
## English Setter 0 0 0
## German shorthaired 0 0 0
## Great Pyreness 0 0 0
## Havanese 0 0 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 0 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 119 0 0
## Samoyed 0 120 0
## Scottish Terrier 0 0 119
## Shiba Inu 0 0 0
## Siamese 0 0 0
## Sphynx 0 0 0
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value Shiba Inu Siamese Sphynx
## Abyssinian 0 0 0
## American bulldog 0 0 0
## American Pit Bull Terrier 0 0 0
## Basset Hound 0 0 0
## Beagle 0 0 0
## Bengal 0 0 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 0 0 0
## Egyptian Mau 0 0 0
## English Setter 0 0 0
## German shorthaired 0 0 0
## Great Pyreness 0 0 0
## Havanese 0 0 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 0 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 0 0
## Shiba Inu 120 0 0
## Siamese 0 120 0
## Sphynx 0 0 120
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value Staffordshire Bull Terrier Wheaten Terrier
## Abyssinian 0 0
## American bulldog 1 0
## American Pit Bull Terrier 2 0
## Basset Hound 0 0
## Beagle 0 0
## Bengal 0 0
## Birman 0 0
## Bombay 0 0
## Boxer 0 0
## British Shorthair 0 0
## Chihuahua 0 0
## Egyptian Mau 0 0
## English Setter 0 0
## German shorthaired 0 0
## Great Pyreness 0 0
## Havanese 0 0
## Japanese Chin 0 0
## Keeshond 0 0
## Leonberger 0 0
## Maine Coon 0 0
## Miniature Pinscher 0 0
## Newfoundland 0 0
## Persian 0 0
## Pomeranian 0 0
## Pug 0 0
## Ragdoll 0 0
## Russian Blue 0 0
## Saint Bernard 0 0
## Samoyed 0 0
## Scottish Terrier 0 0
## Shiba Inu 0 0
## Siamese 0 0
## Sphynx 0 0
## Staffordshire Bull Terrier 114 0
## Wheaten Terrier 0 120
## Yorkshire Terrier 0 0
## prediction
## value Yorkshire Terrier
## Abyssinian 0
## American bulldog 0
## American Pit Bull Terrier 0
## Basset Hound 0
## Beagle 0
## Bengal 0
## Birman 0
## Bombay 0
## Boxer 0
## British Shorthair 0
## Chihuahua 0
## Egyptian Mau 0
## English Setter 0
## German shorthaired 0
## Great Pyreness 0
## Havanese 0
## Japanese Chin 0
## Keeshond 0
## Leonberger 0
## Maine Coon 0
## Miniature Pinscher 0
## Newfoundland 0
## Persian 0
## Pomeranian 0
## Pug 0
## Ragdoll 0
## Russian Blue 0
## Saint Bernard 0
## Samoyed 0
## Scottish Terrier 0
## Shiba Inu 0
## Siamese 0
## Sphynx 0
## Staffordshire Bull Terrier 0
## Wheaten Terrier 0
## Yorkshire Terrier 120
##
## , , = valid
##
## prediction
## value Abyssinian American bulldog
## Abyssinian 72 0
## American bulldog 0 67
## American Pit Bull Terrier 0 3
## Basset Hound 0 0
## Beagle 0 1
## Bengal 4 0
## Birman 0 0
## Bombay 0 0
## Boxer 0 3
## British Shorthair 1 0
## Chihuahua 0 0
## Egyptian Mau 5 0
## English Setter 0 0
## German shorthaired 0 1
## Great Pyreness 0 0
## Havanese 0 0
## Japanese Chin 0 0
## Keeshond 0 0
## Leonberger 0 0
## Maine Coon 1 0
## Miniature Pinscher 0 0
## Newfoundland 0 0
## Persian 0 0
## Pomeranian 0 0
## Pug 0 2
## Ragdoll 0 0
## Russian Blue 1 0
## Saint Bernard 0 0
## Samoyed 0 0
## Scottish Terrier 0 0
## Shiba Inu 0 0
## Siamese 1 0
## Sphynx 0 0
## Staffordshire Bull Terrier 0 5
## Wheaten Terrier 0 1
## Yorkshire Terrier 0 0
## prediction
## value American Pit Bull Terrier Basset Hound Beagle
## Abyssinian 0 0 0
## American bulldog 9 0 0
## American Pit Bull Terrier 59 2 0
## Basset Hound 1 78 1
## Beagle 1 12 61
## Bengal 0 1 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 2 1 0
## British Shorthair 0 0 0
## Chihuahua 0 1 1
## Egyptian Mau 0 0 0
## English Setter 0 0 0
## German shorthaired 1 0 1
## Great Pyreness 0 0 0
## Havanese 0 0 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 2 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 1 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 0 0
## Shiba Inu 0 0 0
## Siamese 0 0 1
## Sphynx 0 0 0
## Staffordshire Bull Terrier 9 1 1
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value Bengal Birman Bombay Boxer British Shorthair
## Abyssinian 3 0 0 0 0
## American bulldog 0 0 0 1 0
## American Pit Bull Terrier 0 0 0 2 0
## Basset Hound 0 0 0 0 0
## Beagle 0 0 0 0 0
## Bengal 68 0 0 0 0
## Birman 0 73 0 0 0
## Bombay 0 0 79 0 0
## Boxer 0 0 0 72 0
## British Shorthair 0 0 0 0 69
## Chihuahua 0 0 0 0 0
## Egyptian Mau 11 0 1 0 0
## English Setter 0 0 0 0 0
## German shorthaired 0 0 0 0 0
## Great Pyreness 0 0 0 0 0
## Havanese 0 0 0 0 0
## Japanese Chin 0 0 0 0 0
## Keeshond 0 0 0 0 0
## Leonberger 0 0 0 1 0
## Maine Coon 2 0 1 0 0
## Miniature Pinscher 0 0 0 0 0
## Newfoundland 0 0 0 0 0
## Persian 0 0 1 0 1
## Pomeranian 0 0 0 0 0
## Pug 0 0 0 1 0
## Ragdoll 0 18 0 0 0
## Russian Blue 0 0 3 0 5
## Saint Bernard 0 0 0 2 0
## Samoyed 0 0 0 0 0
## Scottish Terrier 0 0 0 0 0
## Shiba Inu 0 0 0 0 0
## Siamese 0 16 0 0 0
## Sphynx 0 0 0 0 0
## Staffordshire Bull Terrier 0 0 0 0 0
## Wheaten Terrier 0 1 0 0 0
## Yorkshire Terrier 0 0 0 0 0
## prediction
## value Chihuahua Egyptian Mau English Setter
## Abyssinian 0 1 0
## American bulldog 0 0 0
## American Pit Bull Terrier 1 0 0
## Basset Hound 0 0 0
## Beagle 0 0 1
## Bengal 0 4 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 69 0 0
## Egyptian Mau 0 62 0
## English Setter 0 0 77
## German shorthaired 0 0 0
## Great Pyreness 0 0 0
## Havanese 0 0 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 3 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 0 0
## Shiba Inu 0 0 0
## Siamese 1 0 0
## Sphynx 1 0 0
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value German shorthaired Great Pyreness Havanese
## Abyssinian 0 0 0
## American bulldog 0 1 0
## American Pit Bull Terrier 3 0 0
## Basset Hound 0 0 0
## Beagle 1 1 0
## Bengal 0 0 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 0 0 0
## Egyptian Mau 0 0 0
## English Setter 1 1 0
## German shorthaired 76 1 0
## Great Pyreness 0 80 0
## Havanese 0 0 74
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 0 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 5 0
## Scottish Terrier 0 0 0
## Shiba Inu 0 0 0
## Siamese 0 0 0
## Sphynx 0 0 0
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 1 0 1
## Yorkshire Terrier 0 0 1
## prediction
## value Japanese Chin Keeshond Leonberger Maine Coon
## Abyssinian 0 0 0 0
## American bulldog 0 0 0 0
## American Pit Bull Terrier 0 0 1 0
## Basset Hound 0 0 0 0
## Beagle 0 0 0 0
## Bengal 0 0 0 1
## Birman 0 0 0 0
## Bombay 0 0 0 0
## Boxer 0 0 1 0
## British Shorthair 0 0 0 0
## Chihuahua 0 0 1 0
## Egyptian Mau 0 0 0 0
## English Setter 0 0 0 0
## German shorthaired 0 0 0 0
## Great Pyreness 0 0 0 0
## Havanese 0 0 0 0
## Japanese Chin 80 0 0 0
## Keeshond 0 79 1 0
## Leonberger 0 1 78 0
## Maine Coon 0 0 0 66
## Miniature Pinscher 0 0 0 0
## Newfoundland 0 0 1 0
## Persian 0 0 0 1
## Pomeranian 0 0 0 0
## Pug 0 0 0 0
## Ragdoll 1 0 0 0
## Russian Blue 0 0 0 0
## Saint Bernard 0 0 0 0
## Samoyed 0 0 0 0
## Scottish Terrier 0 0 0 0
## Shiba Inu 0 0 0 0
## Siamese 0 0 1 0
## Sphynx 0 0 0 0
## Staffordshire Bull Terrier 0 0 0 0
## Wheaten Terrier 0 0 0 0
## Yorkshire Terrier 2 0 0 0
## prediction
## value Miniature Pinscher Newfoundland Persian
## Abyssinian 1 0 0
## American bulldog 0 0 0
## American Pit Bull Terrier 1 0 0
## Basset Hound 0 0 0
## Beagle 0 0 0
## Bengal 0 0 0
## Birman 0 0 1
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 7 0 0
## Egyptian Mau 0 0 0
## English Setter 0 0 0
## German shorthaired 0 0 0
## Great Pyreness 0 0 0
## Havanese 0 1 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 4
## Miniature Pinscher 74 0 0
## Newfoundland 0 79 0
## Persian 0 0 73
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 0 2
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 1 0
## Shiba Inu 1 0 0
## Siamese 0 0 0
## Sphynx 0 0 0
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 1 0 0
## prediction
## value Pomeranian Pug Ragdoll Russian Blue
## Abyssinian 0 0 0 0
## American bulldog 0 0 0 0
## American Pit Bull Terrier 0 0 0 0
## Basset Hound 0 0 0 0
## Beagle 0 0 0 0
## Bengal 0 0 0 0
## Birman 0 0 5 0
## Bombay 0 0 0 1
## Boxer 0 0 0 0
## British Shorthair 0 0 1 9
## Chihuahua 0 0 0 0
## Egyptian Mau 0 0 0 1
## English Setter 0 0 0 0
## German shorthaired 0 0 0 0
## Great Pyreness 0 0 0 0
## Havanese 0 0 0 0
## Japanese Chin 0 0 0 0
## Keeshond 0 0 0 0
## Leonberger 0 0 0 0
## Maine Coon 0 0 6 0
## Miniature Pinscher 0 0 0 0
## Newfoundland 0 0 0 0
## Persian 0 0 4 0
## Pomeranian 79 0 0 0
## Pug 0 76 0 0
## Ragdoll 0 0 58 0
## Russian Blue 0 0 0 71
## Saint Bernard 0 0 0 0
## Samoyed 0 0 0 0
## Scottish Terrier 0 0 0 0
## Shiba Inu 0 0 0 0
## Siamese 0 0 1 0
## Sphynx 0 0 0 0
## Staffordshire Bull Terrier 0 0 0 0
## Wheaten Terrier 1 0 0 0
## Yorkshire Terrier 0 0 0 0
## prediction
## value Saint Bernard Samoyed Scottish Terrier
## Abyssinian 0 0 0
## American bulldog 2 0 0
## American Pit Bull Terrier 0 0 0
## Basset Hound 0 0 0
## Beagle 1 0 0
## Bengal 0 0 0
## Birman 0 0 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 0 0 0
## Egyptian Mau 0 0 0
## English Setter 1 0 0
## German shorthaired 0 0 0
## Great Pyreness 0 0 0
## Havanese 0 0 1
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 0 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 1 0
## Pug 0 0 0
## Ragdoll 0 0 0
## Russian Blue 0 0 0
## Saint Bernard 78 0 0
## Samoyed 0 74 1
## Scottish Terrier 0 0 75
## Shiba Inu 0 0 0
## Siamese 0 0 0
## Sphynx 0 0 0
## Staffordshire Bull Terrier 0 1 0
## Wheaten Terrier 0 0 1
## Yorkshire Terrier 0 0 0
## prediction
## value Shiba Inu Siamese Sphynx
## Abyssinian 0 0 3
## American bulldog 0 0 0
## American Pit Bull Terrier 0 0 0
## Basset Hound 0 0 0
## Beagle 0 0 0
## Bengal 0 1 0
## Birman 0 1 0
## Bombay 0 0 0
## Boxer 0 0 0
## British Shorthair 0 0 0
## Chihuahua 1 0 0
## Egyptian Mau 0 0 0
## English Setter 0 0 0
## German shorthaired 0 0 0
## Great Pyreness 0 0 0
## Havanese 0 0 0
## Japanese Chin 0 0 0
## Keeshond 0 0 0
## Leonberger 0 0 0
## Maine Coon 0 0 0
## Miniature Pinscher 0 0 0
## Newfoundland 0 0 0
## Persian 0 0 0
## Pomeranian 0 0 0
## Pug 0 0 0
## Ragdoll 0 1 0
## Russian Blue 0 0 0
## Saint Bernard 0 0 0
## Samoyed 0 0 0
## Scottish Terrier 0 1 0
## Shiba Inu 78 0 0
## Siamese 0 59 0
## Sphynx 0 0 79
## Staffordshire Bull Terrier 0 0 0
## Wheaten Terrier 0 0 0
## Yorkshire Terrier 0 0 0
## prediction
## value Staffordshire Bull Terrier Wheaten Terrier
## Abyssinian 0 0
## American bulldog 0 0
## American Pit Bull Terrier 8 0
## Basset Hound 0 0
## Beagle 1 0
## Bengal 0 0
## Birman 0 0
## Bombay 0 0
## Boxer 1 0
## British Shorthair 0 0
## Chihuahua 0 0
## Egyptian Mau 0 0
## English Setter 0 0
## German shorthaired 0 0
## Great Pyreness 0 0
## Havanese 0 2
## Japanese Chin 0 0
## Keeshond 0 0
## Leonberger 0 0
## Maine Coon 0 0
## Miniature Pinscher 1 0
## Newfoundland 0 0
## Persian 0 0
## Pomeranian 0 0
## Pug 0 0
## Ragdoll 0 0
## Russian Blue 0 0
## Saint Bernard 0 0
## Samoyed 0 0
## Scottish Terrier 1 0
## Shiba Inu 0 1
## Siamese 0 0
## Sphynx 0 0
## Staffordshire Bull Terrier 59 0
## Wheaten Terrier 0 72
## Yorkshire Terrier 0 0
## prediction
## value Yorkshire Terrier
## Abyssinian 0
## American bulldog 0
## American Pit Bull Terrier 0
## Basset Hound 0
## Beagle 0
## Bengal 1
## Birman 0
## Bombay 0
## Boxer 0
## British Shorthair 0
## Chihuahua 0
## Egyptian Mau 0
## English Setter 0
## German shorthaired 0
## Great Pyreness 0
## Havanese 2
## Japanese Chin 0
## Keeshond 0
## Leonberger 0
## Maine Coon 0
## Miniature Pinscher 0
## Newfoundland 0
## Persian 0
## Pomeranian 0
## Pug 0
## Ragdoll 0
## Russian Blue 0
## Saint Bernard 0
## Samoyed 0
## Scottish Terrier 2
## Shiba Inu 0
## Siamese 0
## Sphynx 0
## Staffordshire Bull Terrier 0
## Wheaten Terrier 2
## Yorkshire Terrier 76These images have the highest probabilities of being in a class:
set.seed(1)
par(mfrow = c(2, 3))
y_pred_mat <- predict(model, X)
id <- apply(y_pred_mat, 2, which.max)
for (i in id) try({
par(mar = rep(0, 4L))
plot(0,0,xlim=c(0,1),ylim=c(0,1),axes= FALSE,type = "n")
Z <- image_to_array(image_load(pets$path_to_image[i], target_size = c(224,224)))
rasterImage(Z /255,0,0,1,1)
text(0.5, 0.1, label = class_names[y_pred[i] + 1L], col = "red", cex=2)
})
